Multigroup Analysis with 4 Groups (OTG)

PLS is broadly applied in modern business research. This forum is the right place for discussions on the use of PLS in the fields of Marketing, Strategic Management, Information Technology etc.
Post Reply
PhilipBeu
PLS User
Posts: 15
Joined: Tue May 09, 2017 8:01 am
Real name and title: Philip Beuße

Multigroup Analysis with 4 Groups (OTG)

Post by PhilipBeu » Tue Jul 11, 2017 5:27 pm

Hello everyone,

I´m recently analyzing data for my master thesis.

What I did was a manipulation of 2 specific factors in order to show their influence on behavioral intention. Thus I have 4 groups:

Group1: Factor 1 manipulated
Group2: Factor 2 manipulated
Group3: Factor 1 & factor 2 manipulated
Group4: none of the factors manipulated

What I first did was analyzing wether the manipulation worked (necessary condition) - I did this by ANOVA in SPSS for both factors.

Here starts my first question: Is this correct, or do I have to use FIMIX-PLS?

All 4 SEM-Models look fine. Now we get to the interesting part: the comparison.

In Hair et al. 2017 (The PLS-Book) it says that the Omnibus test of Group Variances proposed in Sarstedt et al. 2011 is the correct way for testing multiple groups. Unfortunately it is not yet implemented in SmartPLS 3.

Is there a easy tutorial like video, excel, etc. which shows an step by step OTG with Data (just like the ones on the homepage for MGA etc.)?

Kind regards,

Philip

User avatar
cringle
SmartPLS Developer
Posts: 797
Joined: Tue Sep 20, 2005 9:13 am
Real name and title: Prof. Dr. Christian M. Ringle
Location: Hamburg (Germany)
Contact:

Re: Multigroup Analysis with 4 Groups (OTG)

Post by cringle » Wed Jul 12, 2017 9:36 pm

You use FIMIX-PLS to uncover unobserved heterogeneity.

Since you know your groups, you do not need to run FIMIX-PLS.

You can compare each group pairwise. When doing such multiple pairwise comparisons, you need to use a correction (e.g., Bonferroni) to account for the multiple testing issue. You find more on this in the book on advanced PLS-SEM issues:
https://www.smartpls.com/documentation/ ... sem-issues

For the OTG, no video exists. This analysis just tells you if there are signficantly differences across multiple groups (e.g. four). But this test does not give you a result if a certain group-specific coefficient is larger than another etc. Hence, the test just tells you if there is something.

You can download the OTG test R code and a data example here: https://www.pls-sem.net/downloads/advan ... pls-sem-1/

Best
Christian

PhilipBeu
PLS User
Posts: 15
Joined: Tue May 09, 2017 8:01 am
Real name and title: Philip Beuße

Re: Multigroup Analysis with 4 Groups (OTG)

Post by PhilipBeu » Thu Aug 03, 2017 2:22 pm

Dear Prof. Ringle,

thank you very much for your advices and your helpful replies (not only in my post but across all of the forum).

If I do the pairwise comparison anyways, for all groups, is the OTG still necessary as a first step or can I leave it (of course be keeping in mind the restrictions of pairwise comparisons)?

Best regards,

Philip

Victoria01
PLS Junior User
Posts: 6
Joined: Wed Dec 20, 2017 11:33 am
Real name and title: Victoria Oers

Re: Multigroup Analysis with 4 Groups (OTG)

Post by Victoria01 » Sun Jan 14, 2018 4:38 pm

cringle wrote: You can compare each group pairwise. When doing such multiple pairwise comparisons, you need to use a correction (e.g., Bonferroni) to account for the multiple testing issue. You find more on this in the book on advanced PLS-SEM issues:
https://www.smartpls.com/documentation/ ... sem-issues
Hi all,

I have a question regarding the pairwise comparison with the permutation approach and the Bonferroni correction. In the book you mentioned, it is described to conduct the permutation test with a significance level of alpha/m (i.e., if you have 5 groups, there would be m=10 comparisons, so you would test a significance level of .005 instead of .05).

Now, am I correct that this means that in SmartPLS, for running pairwise permutation, one enters 0.005 (sticking to the example) for the significance level in the basic settings instead of the preset 0.05?

And, concerning interpretation: how do I evaluate whether the paths differ significantly between the two groups? If I check the significance value, do I base my evaluation on the new significance level (.005) or on the "old", standard 5%?

Thanks a lot for your help!

User avatar
cringle
SmartPLS Developer
Posts: 797
Joined: Tue Sep 20, 2005 9:13 am
Real name and title: Prof. Dr. Christian M. Ringle
Location: Hamburg (Germany)
Contact:

Re: Multigroup Analysis with 4 Groups (OTG)

Post by cringle » Sun Jan 14, 2018 5:10 pm

Hi, you seem to interpret the Bonferroni correction correctly. You would the adjusted probability of error level for your assessment.

In PLS-SEM a similar case is the confirmatory tetrad analysis. Here, to judge if a measurement model is reflective or formative, one needs to look at multiple tests per construct. Here, the Bonferroni correction also changes the probability of error level for the assessment per construct (https://www.sciencedirect.com/science/a ... 6308000143).

Best
CR

Victoria01
PLS Junior User
Posts: 6
Joined: Wed Dec 20, 2017 11:33 am
Real name and title: Victoria Oers

Re: Multigroup Analysis with 4 Groups (OTG)

Post by Victoria01 » Sun Jan 14, 2018 5:25 pm

Great, thanks for your response! :)

Post Reply